Continual Adversarial Reinforcement Learning (CARL) of False Data Injection detection: forgetting and explainability

TL;DR

This paper introduces CARL, a continual adversarial reinforcement learning approach to improve false data injection attack detection in smart inverters, addressing adversarial stealthiness and explainability.

Contribution

It presents a novel continual RL method that incorporates adversarial examples into detection training, highlighting the issue of catastrophic forgetting and proposing a joint training solution.

Findings

CARL enhances detection robustness against adversarial examples.

Joint training mitigates catastrophic forgetting in continual learning.

The approach improves explainability of detection deficiencies.

Abstract

False data injection attacks (FDIAs) on smart inverters are a growing concern linked to increased renewable energy production. While data-based FDIA detection methods are also actively developed, we show that they remain vulnerable to impactful and stealthy adversarial examples that can be crafted using Reinforcement Learning (RL). We propose to include such adversarial examples in data-based detection training procedure via a continual adversarial RL (CARL) approach. This way, one can pinpoint the deficiencies of data-based detection, thereby offering explainability during their incremental improvement. We show that a continual learning implementation is subject to catastrophic forgetting, and additionally show that forgetting can be addressed by employing a joint training strategy on all generated FDIA scenarios.